An outage tracker has an awkward operating constraint: it is needed most at the exact moment the infrastructure around it is degrading. During a power cut in Sri Lanka the home router is down, the local mobile network is strained, and the user’s phone is the one working screen, running on a battery they would rather not drain. The information needed at that moment mostly exists. The Ceylon Electricity Board publishes outage data, and neighbours know precisely what is happening on their street, which is how people distinguish a grid fault from their own tripped main. The two sources never meet, and neither is convenient from a phone during an outage. GridPulse puts both on one map, designed from that failure case outward.
Merging official and community reports
GridPulse reads official outages from the publicly available CEB Care endpoints and displays them beside reports submitted by people currently affected. The map draws each source as a toggleable layer, and the home screen reduces the data to a direct answer: whether the power is out at your location, and how many neighbours are affected.
The main data problem is duplication. If a user reports a cut that CEB has already announced, a naive design renders two markers for one event and the map degrades into noise as reports accumulate. GridPulse merges an official announcement and its matching community reports into a single outage record backed by two kinds of evidence. The project is not affiliated with or endorsed by CEB; it reads what CEB publishes and adds the street-level picture CEB cannot see.
Anonymous reporting as a participation requirement
Filing a report takes two taps: pick the location, optionally add a note, submit. There is no account. On first launch the app assigns a random pseudonym, such as “Brave Peacock”, and that is the only identity attached to anything the user reports. The app collects no email address, no phone number, and no tracking data.
The decision came from participation economics rather than a privacy position. Crowd data is only as good as the crowd, and the crowd only forms when contributing costs nothing; a user sitting in a dark house will not complete an email verification flow to report the obvious. The trade-off is stated plainly: nothing prevents a wrong or bad-faith report. The official CEB layer provides a verified baseline, and merged records give each report more context than it would carry alone, but the community layer runs on good faith. I accepted that risk rather than add friction, because a fully verified map that nobody reports into carries no information.
Offline as the primary state
GridPulse is a PWA that treats connectivity as optional. After the first visit, the app shell, the map tiles for the user’s area, and the last known set of outages are cached on the device, so the app opens and renders with no connection at all. A report filed offline enters a local queue and syncs automatically when the connection returns. The interface follows the same constraint: large touch targets, one-handed layouts, and a dark mode intended to be read in an unlit room.
Resolving location without GPS
GPS is optional. When it is disabled, or when a user is checking on family in another town, a type-ahead search covers every populated place in the country, and a saved home location removes the need to search repeatedly. The geocoding and population data behind that search come from GeoPop, another Prabhava Labs project: a self-hosted Sri Lanka geocoder. A free app therefore does not depend on a paid geocoding API to resolve place names.
Localization down to the name
The app runs in English, Sinhala, and Tamil, with every label and message translated. The name is localized as well: in Sinhala the app is කරන්ට් කට්, literally “current cut”, which is the term the event goes by in daily speech, and the Tamil name, கரன்ட் கட், is the same phrase. The name follows the vernacular term for the event rather than the brand.
Beyond the live view, a stats page reports island-wide totals, active outages and people affected, today compared with yesterday, the worst-hit areas at the moment, and a per-area drilldown with peak-hour history. Once outage data is flowing through the system, the marginal cost of aggregating it is low, and the aggregate view turns individual outages into a pattern.
Status
GridPulse is live at gridpulse-cyr.pages.dev, with source at github.com/prabhavalabs/gridpulse under MIT.
Two known risks remain. The integration depends on publicly available CEB Care endpoints rather than an official API, so it can break whenever CEB changes something on their side; the only available mitigation is adapting quickly. The community layer also has a standard cold-start problem: reports make the map useful, and a useful map attracts reporters, but that loop has to be started by early users reporting into a quiet map.
Moving forward
Two design conclusions carry beyond this project. Building the client for its worst network case produced a better client in every case, and for crowd-sourced data, removing contribution friction matters more than verifying contributors. The design is not specific to Sri Lanka: the MIT license allows forking the project, swapping the official data source, and pointing it at another country’s grid.